互补色小波颜色恒常性/白平衡方法  

Color Constancy With Complementary Color Wavelets

作　　者：陈扬 李旦 张建秋 CHEN Yang;LI Dan;ZHANG Jian-Qiu(Department of Electronic Engineering and the Research Center of Smart Networks and Systems,School of Information Science and Technology,Fudan University,Shanghai 200433)

机构地区：[1]复旦大学信息科学与工程学院智慧网络与系统研究中心和电子工程系,上海200433

出　　处：《自动化学报》2020年第7期1378-1389,共12页Acta Automatica Sinica

基　　金：国家自然科学基金(61571131)资助。

摘　　要：借助于互补色小波,本文提出一种新的颜色恒常性统计方法.分析表明:标准光照图像的互补色小波子带关系,可以利用联合拉普拉斯分布来进行描述.统计学习标准光照图像,可获得拉普拉斯分布的参数,为图像建立起标准光照的基准模型.该基准模型可为光照偏移(颜色恒常偏移)的图像提供光照补偿依据,使偏光图像通过光照补偿恢复为标准光照图像,从而得到光照参数.基于该基准模型对补偿光照参数进行最大似然估计的实验结果表明:本文所提方法的处理效果与列出的最好文献算法相当,其在常用数据库上估计到的光照参数误差中值小0.1°,而均值和最大值则小0.3°.By means of the recent reported complementary color wavelet transform(CCWT),a novel color constancy statistical method is proposed in this paper.Analyses show that the CCWT subband coefficients of the standard light images can be well described by a multivariate Laplace distribution.Learning the distribution parameters from the white light images helps to established the standard multivariate Laplace distribution.Such standard distribution provides the criterion for light compensation,by which the bias light images can be translated to the standard light images.The maximum likelihood estimation results of the compensation light show that:our method is as good as the best performances of listed literatures,reducing the median error 0.1°,and the mean and maximum errors 0.3°.

关 键 词：颜色恒常性 白平衡 互补色小波 彩色图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]